Editorial method

Every claim gets an evidence boundary

ModelRun Lab organizes source material and writes original navigation and explanation. We do not mirror model files, republish source articles, or claim first-hand tests that were not performed.

Level 1

Official

Published by the model maker, repository owner, framework maintainer, or another first-party authority. We still attribute the claim instead of presenting it as our test result.

Level 2

Community

A traceable third-party report. Configuration-specific results are not generalized into universal hardware requirements.

Level 3

Estimated

A calculation based on visible assumptions and a reproducible formula. Estimates are not measurements.

Level 4

Unknown

Used when the available evidence does not support a reliable answer. Unknown is preferable to invented precision.